Electoral debates and campaigns are relevant in providing information and influencing opinion, when assessing political behavior and vote choice. Campaigns and debates have considerable influence over the contours of candidate support, and they are important events that shape the election outcome. For political analysis, the internet offers new means of political access and activity and is more common among young people than older (Dalton 2019, pp. 78). Lately, citizens are more likely to share their political opinions on social media. In 2020, around seven-in-ten adult Twitter users in the U.S. (71%) get news on the site, and 42% U.S. adults on Twitter say they use the site to discuss politics (Pew Research 2020).
This research asks, do pre-existing political opinions affect the way people tweet about their preferred candidates? Why people might change their level of confidence in sharing their opinions on the candidates after a political debate?
In this study, I contend that people, while maintaining their previous political positions, may nonetheless change the way they express their opinions on social media: they adjust their style and tone depending on the candidate performance. This argues against against the conventional wisdom that social media favor polarization. There is extensive literature that argues that social media create echo-chambers where individuals are primarily exposed to like-minded views. This research focuses on the political behavior of Twitter users, and how they communicate their positions about the candidates of 2020 US Presidential election, from the first presidential debate on September 29th to the second debate on November 4th, 2020.
The research question addressed in this study focuses on opinion sharing and opinion shift on social media, and how these behaviors are impacted by party cues and debates. The research question implies causal inference, where the dependent variable is political behavior of Twitter users and the independent variable is their previous political ideologies in their biography or description. The other independent variable is the political debate. This is measured by the most tweeted hashtags referring to the candidates.
The figure above shows a histogram where each bar represents the daily average of tweets according each category. The bar is colored according to the most tweeted category. On average, anti-Trump tweets (in green) is occurring constantly within the two debates, and after the election day on Nov. 4th.
The second frequent category is the one containing hashtags that favor the Democratic party and President Biden (in blue). This is followed by pro-Trump tweets (in purple). While pro-Trump tweets are more present after the first debate, pro-Democratic hashtags become mostly used after the second debate.
Below in Fig.3, I show one word-cloud that plots the most tweeted and frequent words within the two debates and during the Election Day (Nov. 4th).
The word cloud above, contains the most frequent words between the first debate on Sep 27 and the second debate on Oct 22. The most tweeted words were:love, life, proud, news, and Trump. This method allows to see the popularity of words by plotting their frequency. An interesting aspect of this cloud is that it shows that Trump was predominantly cited on tweets more than Biden. This also indicates that Trump might have been tweeted both by affiliated democrats and affiliated republicans.
To understand whether people shifted their direction and attention to the most popular tweets, I selected only the users who tweeted at least twice after the second debate and after the second debate before November 4th, which returned a sample of 282 users. On average, 7% of them are self-identified Trump supporters, 5% Republican, 10% Democrats, and 0.7% anti Trump. For each user, I computed the means of their tweets’ political affiliation for each debate to compare the different political directions.
The graph below shows a stacked bar graph with the distribution of the aggregated values of political affiliations of tweets (y axis) in two time periods (x axis).
The first bar shows a wider distribution of tweets containing hashtags favoring Trump, and less tweets containing tweets that were favoring Biden and the democratic party. The second bar shows a higher percentage of pro Biden and pro Democratic party discourse. Anti-Trump political affiliation of tweets was constant during the two periods.
According to the theories on opinion convergence and bandwagon effect, the graph could suggest that individuals share more their preferences when they are more confident about the candidates’ performances. Accordingly, they might restrain to share their opinions about their initial preferred candidates. Although this is in adherence with the theories discussed above, the measure of popularity of tweets and hashtags is not provided in this study, but only interpreted
In this research I show that people, while maintaining their previous political positions, might change the way they express their opinions on social media as they adjust their style and use of words depending on the candidate performance. In this research, I argue against the conventional wisdom that social media favor polarization, and instead, due the social nature of Twitter, where people not only are exposed to news diets, updated highlights of contentious events like the debates, but they also share their opinions. Although the literature on partisanship suggests that party affiliations affect the direction that people choose to write about their candidates, there are other factors to consider. During electoral debates, partisans might share their opinions favoring their preferred candidates. However, this is unlikely to occur if their preferred candidates are less popular among the other users. On the other hand, people feel more confident about sharing their opinions when they resonate with their preferred candidates’ performances, and their popularity on Twitter.